NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/software/2914

Internet Message Format | 1992-07-22 | 4.0 KB

Xref: sparky comp.software-eng:2914 comp.object:2961
Path: sparky!uunet!mcsun!uknet!axion!srd!news
From: [e-mail] (John Foster)
Newsgroups: comp.software-eng,comp.object
Subject: Re: Software costs revisited?
Message-ID: <[e-mail]>
Date: 22 Jul 92 11:21:00 GMT
References: <[e-mail]>
Sender: [e-mail] (News Administrator)
Reply-To: [e-mail]
Organization: British Telecom
Lines: 76

From article <[e-mail]>, by [e-mail] (Gregor Engelmeier):
> In Brad Cox's book: "Object Oriented Programming" from 1986, he
> quotes an inquiery carried through by the U.S. Goverment Accounting
> Office in 1979, in wich a breakdown of software costs versus actual
> results is performed (FGMSD-80-4). The results at that time were:
>
> Software paid for but not delivered: 47%
> Software delivered but not used:     29%
> Software abandoned or reworked:      19%
> Software used after change:           3%
> Software used as delivered:          2%

Small and pedantic point: the figures of 47% and 29% should be
interchanged there (I have the GAO original in front of me). Sorry --- that
was just for the benefit of anyone who might quote the figures further.

> This was meant to be a motive the introduction of object oriented
> techniques in the production of complex software systems.

I too have heard those figures quoted in that way (in fact I think I've
heard them quoted in support of most new ideas, at one time or another :-)
But on a very quick rescan of the report, I don't think the authors
had that in mind. They were much more concerned with the procedural
and contractual aspects of controlling software development.

Trouble is, those figures on their own don't mean a lot, because they
never were a representative survey. They were found as a result of a large
survey (that was itself representative) but they were just nine specific
cases chosen to illustrate the problems that people were reporting.
They're bound to look bad, because in large part that's why they were
selected! That point is made in the original report, but too rarely
repeated (IMHO).

> My question now is, whether since then such an inquiry has been performed
> again, and which its findings were.

In the light of my comments above, I think I need to re-interpret the
question as: "Has anyone done a survey yielding figures like the above,
but in a way that could be quoted as representative?"
    I have heard of only one such, and I've never seen the original
report. It was supposedly a survey of 125 US companies, carried out
by Applied Research of New Jersey and relating to the percentage of
software developments that were abandoned before delivery. Is there anyone
out there who can supply any detail on that one? I saw it as a comment
in a report itself written in 1987, but it gave neither the date of the
survey nor any reference to a published paper.

Like the poster, I'd very much welcome any further references. But I've
been looking out for this sort of stuff for some years without success,
so I'll be very surprised if we get many (any?) examples.

> Since the OO-Paradigm has been applied predominantly in the design area,
> I would expect, that the relative importance of the first cathegory ("paid
> for but not delivered") has diminished but that the second cathegory of
> failure has become more important.

Sadly, I don't think we're going to be able to answer this: even if we
can find a later data point, the original is unusable as a comparison.
I'd be very happy to be proved wrong on this...

It is, by the way, unfortunate that the original GAO report has become
so famous just on the basis of those particular figures. It's still a
good read, and the mistakes it warns against are still being made...

> I'm greatefull for any hints and comments
>
> - Gregor

Hope this helps.
        John Foster


--
John Foster
Systems Research Division, BT Labs, Martlesham Heath, Ipswich IP5 7RE, UK
[e-mail]        fax: [phone]